Expanding species-specific anatomy ontologies to include the cell ontology

Melissa Haendel (1), Ceri Van Slyke (1), Chris Mungall (2), Peiran Song (1), Oliver Hofmann (3), John Day-Richter (2), ZFIN staff (1)

(1) ZFIN (2) Lawrence Berkeley National Laboratory; (3) South African National Bioinformatics Institute

Abstract

Species-specific anatomy ontologies generally include structures at a gross level. In order to support reasoning down to the cellular level, the cell ontology (CL) may be used in combination with species-specific anatomy ontologies. However, the CL is species-independent and this is problematic because a generic species-independent cell type cannot be part of a specific species. For example, the CL:Purkinje cell is not part of the zebrafish because it represents all Purkinje cells in all organisms in which it exists. We present an ontologically correct method for incorporating the CL into species-specific anatomy ontologies. A tagged subset of the CL (a slim) is generated for the species, and then integrated into the species-specific anatomy ontology. Part_of relations may then be added between the cell types and gross anatomical structures, facilitating annotation and querying.

Synchronizing with the cell ontology

The problem:

A user queries for all zebrafish annotations to cerebellum and expects to get back annotations to the Purkinje cell, a cell type found in the cerebellum.

ZFA:cerebellum CL:Purkinje cell

?

part_of

part_of

Solution:Create species specific slim of the CL and incorporate it into the species’ anatomy ontology

Make a slim by selecting appropriate terms from the CL. •Eliminate cells that don’t exist in fish.•Eliminate extraneous mid-level nodes.

Maintain CL IDs for synchronization.

The OBO-Edit documentation has full description of how to create a slim.

Determining which mid level nodes to eliminate

After removing ‘enucleate erythrocyte’ which is not present in zebrafish, ‘erythrocyte’ only has one child. Parents with only one child can be eliminated.

Not all cell types relevant to zebrafish need to be copied; terms that group cells by function can be left out.

Some cell types with one child are retained because it is known that more siblings exist that are not yet in the CL.

CL ZFAneuron CL:540 primary neuron CL:530 primary motor neuron

CL:533

neuron ZFA:9248 primary neuron ZFA:9242 primary motor neuron

ZFA:9244 CaP motoneuron

ZFA:6000 VaP motoneuron

ZFA:6001 RoP motoneuron

ZFA:6002 MiP motoneuron

ZFA:60004

Species-specific cell types are added as children of terms with a xref to a CL term. This works as long as they have no children with xrefs to CL.

Species-specific cell types

xref

Using the CL and xrefs from ZFA to CL, OBOEdit can tell that nucleate erythrocyte should be placed under blood cell. The dotted lines indicate links that have been implied by the reasoner.

Cells within the ZFA are species specific subtypes of those in the CL, and the hierarchies should be consistent. The OBOEdit reasoner can use the mapping information to detect missing links in the more specific ontology.

In the ZFA, nucleate erythrocyte is a direct child of cell.

ZFA

In the CL, nucleate erythrocytes are types of blood cells.

Erythrocyte is not in the ZFA, as it is redundant with nucleate erythrocyte. However, nucleate erythrocyte should be is_a blood cell in the ZFA.

CL

xref

Using the OBOEdit reasoner to identify implied links

The OBO Edit2 "assert implied links" menu item uses the reasoner to help the curator make links in ZFA that are consistent with CL. The editor can check the links they believe and save the file, making ZFA logically consistent with CL.